Navigation Sensor Data Reliability Model-Based on Self-Evaluation and Mutual Evaluation

Multi-source navigation involves using multiple available sensors to achieve high-precision location services, but navigation devices and data are susceptible to various environments and attacks, emphasizing the need for pervasive security measures like data reliability evaluation. Common reliability estimation methods for Internet of Things data are weak for highly dynamic navigation data, which are invalid when sensor motion is considered. We proposed a method combining self-evaluation and mutual evaluation to assess the reliability of real-time navigation sensor data. Our approach introduces an innovative Transformer-based model for numerical-based sensors and a feature point prediction consistency test for digit image-based sensors. We also proposed a reconstruction consistency check method for mutual evaluation of multiple digital images. By combining self-assessment and mutual assessment, we determine the final reliability of the sensor data. To enable mutual evaluation of multiple digital images, we proposed a reconstruction consistency check method. By combining the results of self-assessment and mutual assessment, we determined the final reliability of the sensor data. The performance of our method was evaluated using both trustworthy and untrustworthy data. We have observed significant improvements in average absolute errors for UWB positioning and visual odometry positioning through experimentation, utilizing reliability-based weighted fusion, resulting in a reduction of 0.28m for UWB positioning and 0.044m for visual odometry positioning.